This De-escalation Roleplay simulator uses an LLM (large language model) to generate the virtual character’s responses, guide the direction of the conversation, and write personalised feedback for the learner. Furthermore, it is designed around a sensitive subject matter, where aggressive responses are part of the character’s intended behaviour.
We have designed the prompts around the learning outcomes and goals of this simulator, and with strict conversation boundaries that prevent unintended or inappropriate use.
However, please be aware of the following limitations:
- Due to the large nature of its dataset, the LLM is prone to ‘hallucinate’, meaning it may generate information that appears factual but is incorrect, misleading, or entirely fabricated.
- It is impossible to predict everything that users may feed into the customisation fields and everything that they may say in the conversation. Therefore, unintended or misguided uses of the conversation customisation, deliberate attempts to manipulate the system, and unanticipated responses from the learner may cause unexpected outputs from the LLM that are beyond our control.
In short, please be aware that we cannot guarantee all AI-generated content will be accurate, appropriate, or aligned with educational objectives in all circumstances.
We appreciate any feedback about the performance of our simulators. We work hard to design and improve our LLM-based roleplays to give you a personalised, yet safe and impactful learning experience.
.png?width=240&height=350&name=Eric%20(1).png)
.png?width=240&height=350&name=Joe%20(1).png)
